A SUBSPECIFIC REVISION OF NORTH AMERICAN SALTMARSH MALLOW, KOSTELETZKYA PENTACARPOS (L.) LEDEB.

Transcription

1 A SUBSPECIFIC REVISION OF NORTH AMERICAN SALTMARSH MALLOW, KOSTELETZKYA PENTACARPOS (L.) LEDEB. by Sara Nazanin Alexander A Thesis Submitted to the Graduate Faculty of George Mason University in Partial Fulfillment of The Requirements for the Degree of Master of Science Environmental Science and Policy Dr. Andrea Weeks, Thesis Director Dr. Laurence DOff, Committee Member Dr. Robert Jonas, Committee Member Dr. Robert Jonas, Department Chairperson Dr. Richard Diecchio, Associate Dean for Academic and Student Affairs, College of Science Dr. Vikas Chandhoke, Dean, College of Science Date: Summer Semester 2010 George Mason University Fairfax, VA

2 A subspecific revision of North American saltmarsh mallow, Kosteletzkya pentacarpos (L.) Ledeb. A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science at George Mason University By Sara Nazanin Alexander Bachelor of Science College of William and Mary, 2003 Director: Dr. Andrea Weeks, Assistant Professor Department of Environmental Science and Policy Summer Semester 2010 George Mason University Fairfax, VA

3 ACKNOWLEDGMENTS I would foremost like to thank Dr. Andrea Weeks, my adviser, for providing invaluable support and advice through every step of this thesis, whether offering ideas on campus, timely words of encouragement over from the other side of the world, careful attention through multiple edits, or directions on how to set up a tent after a day in the field. Thanks to Dr. Larry Dorr and Dr. Robert Jonas for serving on my committee and providing guidance, whether on practical considerations or the finer points of Malvaceae morphology, and to Dr. Howard Lasus, Dr. Christian Jones, and Chris Ruck for providing essential guidance about statistics. I would also like to thank Annaliesa Guilford for her patience and help in clearing the path toward graduation, and Shannon Granville for proofreading. Dr. Vicki Funk, Dr. Carol Kelloff, and Dr. Gerald Guala at the Smithsonian Institution were generously available for encouragement or experience-based advice. The research of Dr. Orland Blanchard and Dr. Jack Gallagher provided much inspiration and information for this thesis, and corresponding with them was a pleasure as well as a source of vital information. Finally, I always owe my parents a debt of gratitude for their support. ii

4 TABLE OF CONTENTS Page Abstract...v 1. Introduction...1 Taxonomic History of Kosteletzkya pentacarpos...2 Varieties of Kosteletzkya pentacarpos...6 Recognition of Varieties of Kosteletzkya pentacarpos...10 Goals Methods...12 Loans and Collections...12 Determination of Specimens to Varietal Names...13 Measurements...15 Georeferencing...19 Statistical Analyses Results...26 Character Selection...26 Principal Components Analysis...29 Discriminant Analysis...31 Analysis of Variance by Geography Discussion...34 Classification of Infraspecific Variation...34 Recognizing Infraspecific Variation...37 Infraspecific Variation in Kosteletzkya pentacarpos...42 Variety smilacifolia...45 Future Research...49 Appendix A: Tables...51 Sources of loans, total specimens received on loan, and total specimens measured.51 Characters measured...52 Qualitative values key...55 iii

5 Data coverage; characters present for specimens measured...56 Appendix B: Figures...57 Appendix C: Summary of distinguishing characteristics of Kosteletzkya pentacarpos varieties...78 Appendix D: Exsiccate...83 References...89 Curriculum Vitae...96 iv

6 ABSTRACT A SUBSPECIFIC REVISION OF NORTH AMERICAN SALTMARSH MALLOW, KOSTELETKZYA PENTACARPOS (L.) LEDEB. Sara Nazanin Alexander, M.S. George Mason University, 2010 Thesis Director: Dr. Andrea Weeks Kosteletzkya pentacarpos (L.) Ledeb. (syn. = K. virginica (L.) K. Presl ex A. Gray), has been treated as a single variable species or as a species including four varieties: K. virginica var. althaeifolia, var. aquilonia, var. smilacifolia, and var. virginica. The aim of this study was to test the validity of formally recognizing these putative infraspecific taxa by using statistical analysis of morphological data. Loans were received from 15 herbaria located along the east and Gulf coasts of the United States, a total of 1,114 specimens. Measurements were taken of 224 specimens for 43 characters and analyzed using principal components analysis, discriminant analysis, and analysis of variance. Putative varieties althaeifolia and aquilonia are not morphologically or geographically distinct from var. virginica but represent extremes in variation along a continuous latitudinal cline. Plants with dense and long pubescence, biserrate leaf margins, small seeds, and tightly constricted capsules tend to occur in the southern part of the range, while plants with short and sparse pubescence, serrate leaf margins, large seeds, and shallowly

7 constricted capsules occur in the northern part of the range. Variety smilacifolia, however, occurs only in peninsular Florida, and is morphologically distinct from the other taxa in having linear hastate leaves with reflexed linear lobes, an unbranched inflorescence, a slender stem, and a mostly entire leaf margin. A new combination is proposed: Kosteletzkya pentacarpos (L.) Ledeb. var. smilacifolia (Chapm.) S.N. Alexander. All other former putative infraspecific taxa are synonymous with Kosteletzkya pentacarpos var. pentacarpos.

8 Introduction Kosteletzkya pentacarpos (L.) Ledeb. (syn. = K. virginica (L.) K. Presl ex A. Gray), known as the seashore mallow or saltmarsh mallow (Figures 1, 2), is commonly found along the Gulf and east coasts of the United States from Texas and Florida to New York (Figure 3). It also includes populations in Bermuda and Cuba, and Old World populations scattered through Spain, Corsica, Italy, Iran, Azerbaijan, Georgia, and Russia. Because it appears in few scattered locations in the European Union, it is designated as an Annex II species (deserving special areas of conservation) under the Habitats Directive (Pino and de Roa, 2003). In the United States, it is generally limited to the Coastal Plain physiographic region (Figure 4). It is found in brackish to almost freshwater marshes, shores, swamps, and low ground, although certain representatives have been found in low pine woods (K. pentacarpos var. smilacifolia; Chapman, 1883). The flowers are similar to cultivated Hibiscus flowers in their appearance, and their beauty earned the saltmarsh mallow the status of Wildflower of the Year from the North Carolina Botanical Garden in 1990 (NCBG, 2008) and from the Virginia Native Plant Society in 2004 (VNPS, 2008). Kosteletzkya pentacarpos has entered the public eye in recent years as a potential source of biodiesel fuel. Much of the research about the use of K. pentacarpos as a biofuel crop originated with the work of John Gallagher at the University of Delaware. In 1

9 2006 his team issued a patent for the use of mallow as a source of biodiesel (Gallagher et al., 2006). A body of research has begun to accumulate indicating that seashore mallow is potentially a practical, effective, and ecologically sound source of biodiesel. Its seeds yield oil that is comparable to soybean oil in quantity and quality and can be used to make biodiesel with a high cetane number that burns cleanly (Ruan et al., 2008). Unlike soybeans, K. pentacarpos can be cultivated on marginal, salty lands, such as exposed mud flats, irrigated agricultural land in dry climates, or among coastal vegetation that has been killed by seawater during high spring tides (Schill, 2007). Experimental plots of K. pentacarpos cultivated over 10 years in saline mud flats in Jinhai, China, suggest that K. pentacarpos can be cultivated successfully outside its native range without high risk of becoming invasive and so could potentially support small biodiesel industry (Ruan et al., 2008; He et al., 2003). Ecological and agricultural research on K. pentacarpos continues to be conducted at the University of Delaware under the guidance of Dr. Gallagher. Taxonomic History of Kosteletzkya pentacarpos Until recently, New World members of Kosteletzkya pentacarpos were placed in K. virginica (L.) C. Presl. ex A. Gray. Blanchard (2008) reduced the name K. virginica to synonymy under K. pentacarpos, explaining that representatives from the northeastern United States are remarkably similar to Eurasian representatives in most characters including floral features; fruit color, pubescence, and size; seed size, shape, surface features, and color; and overall pubescence of the herbage. While Blanchard noted some 2

10 variation in leaf outline, convexity of the capsule wall, and petal conformation among living specimens grown from seeds from Iran, Italy, and the United States (from New Jersey), he did not find these differences to be significant in the context of the variation the species displays across its North American range (personal communication, 24 September 2008). His conclusions are based on years of observation, but he has not conducted a systematic specimen-based analysis of the differences between the two putative species. Both Kosteletzkya virginica and K. pentacarpos can be traced back to publication as species of Hibiscus in Species Plantarum (Linnaeus, 1753). Linnaeus described both briefly, stating that Hibiscus virginicus occurs in Virginia (lectotype Clayton 567, designated by Reveal in Jarvis, 2007:571), and H. pentacarpos in Venice (type not designated). Linnaeus described H. pentacarpos as Planta antecedenti simillima; sed fructa diversa similar to the preceding plant (H. virginicus), but the fruits are different. However, he did not describe the fruits of either species. His description of H. pentacarpos further differs from his description of H. virginicus in that the leaves of H. virginicus are trinerved and subtomentose, while the leaves of H. pentacarpos are subtrinerved and tomentose. Although both names were published simultaneously in Species Plantarum, Hibiscus pentacarpos has priority over H. virginica. Cavanilles (1787), considered Hibiscus virginicus to be a synonym of H. pentacarpos. Cavanilles noted that Plukenet (1691, pl. 6, fig. 4) presented H. virginicum as different from H. pentacarpos, but wrote Plukenetii vero synonymum reieci utpote diversum a nostro Pentacarpo (Truly, as you 3

11 might expect, I rejected the different synonym of Plukenet for our Pentacarpos). According to the International Code of Botanical Nomenclature (McNeill et al., 2006; Art. 11.5), the name K. pentacarpos must be accepted while K. virginica is relegated to synonymy, because while both basionyms were published simultaneously Cavanilles choice to retain the epithet pentacarpos gives it priority. Later publications also recognized H. virginicus and H. pentacarpos as separate species. Hibiscus virginicus was transferred to the genus Kosteletzkya by Presl in Gray (1849). Presl described the genus as a whole rather than describing any species individually. He said of K. virginica that it is the only species of Kosteletzkya found in the United States, common on the coast from Virginia southward, and is sparingly found as far north as Long Island. His publication also included an illustration of K. virginica, showing the flower and fruit in detail. Carl Friedrich von Ledebour (1842) proposed the combination Kosteletzkya pentacarpa (L.) Ledeb. Because the generic name Kosteletzkya is feminine, he published the specific epithet in its feminine form, pentacarpa. However, Linnaeus capitalized the epithet in his original publication, and so it is correctly considered to be a generic noun and should always keep the Greek masculine ending (Blanchard, 2008). Kosteletzkya virginica and K. pentacarpos have been considered to be members of the genus Pavonia. Sprengel (1826) proposed the names P. virginica and P. venata in the 16 th edition of Linnaeus Systema vegetabilium. The name P. venata was based on material from Hibiscus pentacarpos L., but because the rules of nomenclature require that a new combination must retain the species epithet, the name P. venata is illegitimate. 4

12 Pavonia pentacarpos (L.) Poir. was published by Poiret (1816) in Lamarck s publication Encyclopédie Méthodique Botanique, Supplément. Pentagonocarpus is another generic name that has been applied to Kosteletzkya virginica and K. pentacarpos. In Flora Italiana, Filipo Parlatore (1873) classified seven Kosteletzkya species as Pentagonocarpus species. He proposed a new name for Hibiscus pentacarpos: Pentagonocarpus zannichellii. However, by the rules of nomenclature this name, like Sprengel s Pavonia venata, is illegitimate. All Pentagonocarpus species were later relegated to synonymy under Kosteletzkya. Plants in the genus Kosteletzkya resemble members of the genus Hibiscus in floral morphology but differ mainly in that each carpel contains a single seed as opposed to multiple seeds (Bayer and Kubitzki, 2003). When proposed by Presl (1831) the name Kosteletzkya applied only to New World species but was later expanded to include Old World species including K. pentacarpos. The genus now includes 17 species (Fryxell, 1988) distributed in Europe, mainland Africa, Madagascar, the Caribbean, and Central and South America, with the largest number of species found in Mexico. Besides K. pentacarpos, only one other species of Kosteletzkya is found in the United States: Kosteletzkya depressa (L.) O.J. Blanch. et al., which has been collected in Texas (Jones et al., 1991). The genus Kosteletzkya C. Presl belongs to subfamily Malvoideae and tribe Hibisceae within Malvaceae s.s. (Bayer and Kubitzki, 2003), that also includes the genera Hibiscus and Pavonia. Molecular phylogenies based on chloroplast and nuclear DNA (Pfeil and Crisp, 2005) have placed K. virginica in the Trionum clade, a lineage that also 5

13 includes the genera Pavonia, Malvaviscus, and Abelmoschus. Species of Kosteletzkya from Madagascar were found to belong to a different clade, suggesting that the genus Kosteletzkya, as defined by Fryxell (1988), is polyphyletic (Koopman and Baum, 2008). However, this polyphyly has not yet been reflected in nomenclatural changes. Because the type of Kosteletzkya, K. hastata K. B. Presl, is a synonym of the New World species K. depressa, New World Kosteletzkya would retain the generic name. Varieties of Kosteletzkya pentacarpos Taxonomists have recognized four varieties of Kosteletzkya pentacarpos: variety althaeifolia Chapm., variety smilacifolia Chapm., variety aquilonia Fernald, and the typical variety, var. virginica. A summary of varietal differences is listed in Appendix C. Variety smilacifolia occurs only in Florida, variety althaeifolia occurs in Virginia and to the south, variety aquilonia occurs in Virginia and to the north, and variety virginica occurs in Delaware and to the south. Two varieties, althaeifolia (as altheaefolia ) and smilacifolia, were described by Alvan Wentworth Chapman (1860) in his Flora of the Southern United States. He distinguishes them by leaf shape, vegetative pubescence, and inflorescence density. Chapman (1860) described the leaves of var. althaeifolia as ovate or ovate-lanceolate, whereas he described the leaves of var. smilacifolia as hastate with lanceolate-serrate lobes. Variety althaeifolia is densely stellate-pubescent and somewhat hoary, while var. smilacifolia is smoothish in its lower parts. Floral racemes are dense in var. 6

14 althaeifolia, and few-flowered in var. smilacifolia. Chapman also noted that the capsule of var. althaeifolia is hirsute. Chapman based the names Kosteletzkya virginica var. althaeifolia and var. smilacifolia on the names Hibiscus althaefolius and H. smilacifolius, which were used by Shuttleworth in Gray (1852). Shuttleworth attempted to describe H. althaefolius and H. smilacifolius as new species in this publication, and the specimens he examined were clearly members of what we now consider Kosteletzkya. However, he did not publish these species validly. Although he designated type specimens Rugel 102 and Rugel 103 as Hibiscus (Pentaspermum) althaefolius and Hibiscus (Pentaspermum) smilacifolius, his designation did not include a description for either taxon. Consequently the names H. althaefolius and H. smilacifolius are nomena nuda and Shuttleworth is not listed as an author for the varietal epithets used by Chapman. A third variety, Kosteletzkya virginica var. aquilonia, was published by Fernald (1941). He described the distribution of variety aquilonia as in the northern part of the range of Kosteletzkya pentacarpos range, from New York to Virginia. He stated that the illustration that accompanied Gray s publication of Presl s K. virginica represented the southern variety virginica, which he called var. typica, which ranges from Florida to the southeastern United States. He also stated that variety althaeifolia (as altheaefolia ) is found in Virginia but did not give its range. Fernald distinguished variety aquilonia from variety virginica and variety althaeifolia by leaf shape, flower size, and flower pubescence, and included a photograph comparing flower size and capsule pubescence of the three. Leaves of variety aquilonia 7

15 are mostly rotund to ovate with narrowly ovate to hastate leaves appearing only at the very top of the stem. The leaves of variety virginica are mostly narrowly ovate to hastate except for the lowest leaves, and the leaves of variety althaeifolia often do not have divergent basal lobes and are not hastate. Variety aquilonia has distinctly smaller flowers (petals 1.8 rarely 3 cm long and 1 2 cm broad) than varieties virginica and althaeifolia (petals cm long). The capsules of variety aquilonia are sparsely hispid-setose with hairs mm long, while the capsules of variety virginica are villous-hirsute with hairs mm long, and the capsules of variety althaeifolia are very heavily villous-hirsute. Variety althaeifolia is further distinguished from the other two varieties by having copious rough tomentum on stem and foliage giving a paler and plush-like appearance, a hispid-tomentose calyx, and thick pedicels that are often shorter than their subtending bracts instead of thin pedicels that are longer than their subtending bracts. Varieties althaeifolia and smilacifolia have each at one time been considered to be species distinct from Kosteletzkya virginica. Variety althaeifolia was raised to species level by Watson (1878). Variety smilacifolia was raised to species level by Chapman (1883), who described the capsule as hirsute; but because he described the capsule of variety althaeifolia as hirsute in his 1860 publication, the characteristic does not distinguish the varieties. Chapman also stated that variety smilacifolia is found in low pine woods in South Florida. A white form of Kosteletzkya pentacarpos var. aquilonia was described by Clyde Reed (1951); K. virginica var. aquilonia forma alba. Collected in Dorchester and Anne 8

16 Arundel counties in Maryland, he described this form as differing from var. aquilonia only in flower color. Reed stated that, the rest of the characters best fit those of variety aquilonia, in having a flowering calyx 6 to 10 mm high, and the bractlets about 5 to 6 mm long, with petals up to 3 cm long and the pedicels of the flowers equaling or longer than the bracteal leaves. Plant taxonomists do not commonly recognize the subspecific division forma because this term may denote sporadic environmental or genetic variation as opposed to the consistent patterns of morphological variation that distinguish species, subspecies, or varieties. 9

17 Recognition of varieties of Kosteletzkya pentacarpos in floras The Flora of North America North of Mexico (volume 6, in review as of April 2010) will consider Kosteletzkya pentacarpos to be one species without varieties. The author of the treatment of Kosteletzkya, Orland Blanchard, acknowledges the variability in K. pentacarpos, but says I distinguish no infraspecific taxa formally, my conclusion being that while there may be some clinal variation in some characters, they do not vary together, with the result that no useful boundaries can be discerned (personal communication, 25 February 2008). Other regional floras have also declined to recognize varieties of Kosteletzkya pentacarpos. The Manual of the Vascular Flora of the Carolinas (Radford et al., 1968) lists K. althaeifolia and K. virginica var. althaeifolia as synonyms of K. virginica, and notes that the species is very variable in leaf shape, pubescence and flower size. Several varieties have been recognized, but the intergrading forms are so numerous as to make separation very difficult. Aquatic and Wetland Plants of the Southeastern United States (Godfrey and Wooten, 1981) also considers K. althaeifolia as well as K. smilacifolia to be synonyms of K. virginica, and explains that members of the species exhibit wide yet intergrading variations in height, pubescence, leaf size and shape, and corolla size. A Flora of Tropical Florida (Long and Lakela, 1971) distinguishes K. althaeifolia from K. virginica based on leaf shape and calyx pubescence, but notes possibly K. althaeifolia is only a variety of K. virginica rather than a separate species. 10

18 However, the Flora of the Carolinas, Virginia, Georgia, northern Florida, and Surrounding Areas (Weakley, 2008) distinguishes between variety virginica and variety aquilonia. It characterizes them by the size of floral parts and length of fruit pubescence, and also by range, as the typical variety occurs in Delaware and further south while variety aquilonia occurs in Virginia and further north. Goals A Flora of Virginia is currently in preparation with an estimated publication date of This will be the first Flora of Virginia to be published since 1743 and will be more comprehensive than the former Flora (Flora of Virginia website, 2008). The initial treatment has followed Alan Weakley s regional flora (Weakley, 2008) in tentatively recognizing varieties althaeifolia and aquilonia (J.F. Townsend, personal communication, February 18, 2010), but the final decision had not been made as of February This infraspecific revision of Kosteletzkya pentacarpos will use systematic specimen-based statistical techniques to decide whether varieties should be formally recognized, as in Weakley s Flora (2008), or not formally recognized, as in Orland Blanchard s treatment in the Flora of North America North of Mexico. The results will influence the treatment of the species in the upcoming Flora of Virginia. 11

19 Methods Loans and collections Loans of Kosteletzkya pentacarpos specimens were received from 15 herbaria located along the east and Gulf coasts of the United States, especially those with a focus on local Coastal Plain flora or on Malvaceae (Table 1). These loans were stored and studied at the herbarium at George Mason University in Fairfax, Virginia. A preliminary table of all 1,114 specimens received was built in a Microsoft Access database. The database included label information and indicated specific specimens that would be useful for analysis. The ideal specimen for analysis would have intact flowers, fruits, and leaves, as well as accurate information on the height of the plant, its habitat, and its locality. Since few specimens are ideal, specimens were selected with the intention of minimizing the amount of missing data. In August 2009, my graduate advisor Andrea Weeks, a fellow graduate student Kristen Baird, and I made a collecting trip to Maryland and Delaware. We collected 23 specimens of Kosteletzkya pentacarpos, 12 of which I measured for analysis. I determined all of these to be variety aquilonia according to the diagnosis in Weakley s Flora (2008), which is based on Fernald s diagnosis in his 1941 publication of variety aquilonia: they all bore sepals less than 10 mm long and epicalyx bracts less than 6 mm 12

20 long, flowers with petals less than 3 cm in length and 2 cm in width, and/or fruits with sparse hairs less than 1.0 mm long. A voucher for each of these collections was mounted and deposited at the George Mason herbarium (GMUF). Determination of specimens to varietal names As a result of the taxonomic history of Kosteletzkya pentacarpos, some floras have not recognized varieties, including the Manual of the Vascular Flora of the Carolinas (Radford et al., 1968) and Aquatic and Wetland Plants of the Southeastern United States (Godfrey and Wooten, 1981), and many specimens were collected by scientists who do not choose to recognize the varieties. As a result, it is practically impossible to know from the label alone whether a determination of Kosteletzkya virginica means that the plant has been confidently assigned to the typical variety or whether it is simply not assigned to any variety. Few specimens had been collected or annotated recently enough to be labeled as pentacarpos. Weakley s Flora (2008) provides the most recent diagnosis between variety aquilonia and variety virginica. It describes the petals of the typical variety as being cm long and 2 3 cm broad, significantly bigger than the petals of var. aquilonia at cm long and 1 2 cm broad. Weakley also states that the sepals of var. aquilonia are mostly 6 10 mm long and the epicalyx bracts mostly mm long, while the sepals of var. virginica are mostly 8 13 mm long and the bracts mostly 6 10 mm long (Appendix C). 13

21 Weakley does not diagnose variety althaeifolia, but according to Fernald (1941), var. althaeifolia is distinguished from var. virginica by having copious rough tomentum on stem and foliage giving a paler and plush-like appearance. Other floras agree with this criterion. The flowers of var. althaeifolia are large, but within the range given for var. virginica, and so not a robust diagnostic character: they are 4 cm long as opposed to 3 4 cm long (Long and Lakela, 1971); about 4 cm long as opposed to 2 4 cm long (Small and Kunkel, 1903). Neither of these treatments diagnose variety smilacifolia. Steere (1968) describes the leaves of variety smilacifolia as being cleft into three narrow lobes, while the leaves of variety virginica and althaeifolia are respectively ovate or angled or lacking angles, and the foliage of var. smilacifolia is smooth. Small and Kunkel (1903) described var. smilacifolia as being glabrous or barely puberulent, and its leaves as being linear and reflexed. Chapman s original publication of the variety describes its leaves as hastate 3-lobed (his 1860 publication says all hastate, his 1883 publication of the variety says the lowest ovate, all others hastate-3-lobed ). All three treatments agree that var. smilacifolia is found in Florida. Therefore, I classified specimens with petals cm long and 1 2 cm broad and short, sparse hairs on the fruit (Table 3) as variety aquilonia, specimens with dense, long pubescence on abaxial and adaxial leaf surfaces, upper stem, and fruit (Table 3) as variety althaeifolia, and specimens from Florida with sparse pubescence and narrow, almost linear hastate leaves as variety smilacifolia, and let others remain as variety virginica. 14

22 There was some uncertainty in this approach. Six specimens had both the heavy pubescence of variety althaeifolia and the small flowers of variety aquilonia, two of which had been determined as var. althaeifolia when collected, and three of which had been determined as var. virginica. I let these remain as they had initially been identified. Given these criteria, 36 specimens labeled as Kosteletzkya virginica were determined to be variety aquilonia, 19 were determined to be var. althaeifolia, and one was determined to be var. smilacifolia. One specimen that was labeled as Kosteletzkya virginica smilacifolia was determined to be K. virginica, as its leaves were moderately pubescent rather than glabrous or smooth, as well as nearly lanceolate rather than cleft into three narrow lobes ; in addition, it was collected in Mississippi. Measurements Specimens were measured for 44 different characters (Table 2). The 44 characters chosen were primarily based on diagnostic descriptions in floras, in which the characters most consistently used are leaf shape, flower size (including epicalyx, calyx, petals, and column), and pubescence (both density and length of hairs) (Appendix C). Other possible diagnostic characters have been used in some treatments: for example, Small and Kunkel (1903) diagnosed variety althaeifolia from varieties virginica and smilacifolia in part by stating that its seeds are glabrous and striate while the seeds of the other two varieties may be wooly. In some cases, several measurements were taken to represent one aspect of the plant: inflorescence length, number of flowers, and number of inflorescence 15

23 branches together served as a proxy for the density of the inflorescence (a character used by Chapman (1860)); eight measurements were taken to represent the size and shape of the leaves. Other characters that seemed intuitively important were also measured, including seed length and width, and the diameter of the stem at the base of the inflorescence. Qualitative characteristics were defined according to Harris and Harris (2001). Qualitative characters with only two states including epicalyx bract shape and sepal tip shape were assigned a 1 or a 0 (Table 3). Vegetative pubescence and fruit pubescence proved to be practically impossible to measure in a precise quantitative manner: hair length was variable, and the stellate shape of the hairs caused them to overlap and blend visually when they were dense, and it became extremely difficult to distinguish individual hairs. Instead, vegetative and fruit pubescence were classified into two threepoint numeric scales each, one for hair length and one for hair density (Table 3; Figure 5). Pubescence on the abaxial surfaces of leaves includes sparse three-armed stellate hairs that are about three times the length of the denser five-armed stellate hairs; the smaller, more common hairs were used in the measurements (Figure 6). A ruler was used to measure larger characters, including leaf and petal dimensions, to the nearest millimeter. The length of the inflorescence was measured to the nearest half-centimeter. A dissecting microscope with a reticule eyepiece with precision of 0.25 mm was used to estimate the length and density of hairs, and a microscope stage-micrometer gauge was used to measure smaller characters, including sepal and seed dimensions, to the nearest tenth of a millimeter. 16

24 In his 1941 publication of variety aquilonia, Fernald stated that for variety althaeifolia, the thick pedicels rarely [equal] their subtending leaves (in vars. typica and aquilonia the more slender pedicels frequently [overtop] the subtending leaves) Because he used this as a diagnostic character, the diameter of the pedicel bearing the flower or fruit being measured from each specimen was measured. A cross-section was cut and measured with a dissecting scope with a gauge. The length of the longest pedicel and the length of the pedicel of the most mature flower (or fruit) were also measured, as were the lengths of their subtending leaflets. The five-angled fruit capsules of K. pentacarpos vary in degree of constriction, and this variation was used as a diagnostic character by Long and Lakela (1971). Degree of capsule constriction was represented by measurements of the capsule both at its widest point (between two carpel points) and narrowest (between two constrictions) (Figure 7). Seed capsule pubescence was classified into two three-point scales for length and density (Table 3; Figure 7). Leaf margin was classified as either serrate or biserrate (Figure 8). The degree of dentation was represented by measuring the distance occupied by five teeth at about the midpoint of the leaf. While leaf shape has been used as a diagnostic character in floras, imprecise terms are often employed. For example, Fernald (1941) described the leaves of variety virginica as usually entire and narrow, althaeifolia as all undivided ovate or ovate-lanceolate, and smilacifolia as all hastate with lanceolate lobes. Steere (1968) described the leaves of var. virginica as ovate or angled, often with indented base, althaeifolia as lacking 17

25 angles, and smilacifolia as cleft into three narrow lobes. To quantify the leaf shape, I took a number of measurements in addition to the length of the petiole, length of the blade, and overall width of the leaf; by measuring many dimensions I intended to express the terms hastate, lanceolate, and palmately lobed in quantities rather than qualities (Table 2; Figure 8, Figure 9). I represented lanceolate leaves lack of lobes by giving them a lobe length 0 and also an angle between the lobe and the central vein of 0. However, for the lanceolate leaves I estimated the length of petiole to sinus and length of sinus to sinus by estimating an angle of 45 degrees (chosen by observation of leaves with lobes, and also by averaging lobe-to-central vein angle of greater than 90 degrees), and treating that point along the leaf s edge as its sinus for the purpose of those measurements. Leaves on a Kosteletzkya plant differ in shape according to their position on the stem, and this difference has been used as a diagnostic character. Fernald (1941) distinguished variety aquilonia from variety virginica in part by stating that the leaves of var. aquilonia are narrowly ovate to hastate only at the very top of the stem, while the leaves of var. virginica are mostly narrowly ovate to hastate except for the lowest leaves. For the sake of consistency, the uppermost leaf below the inflorescence was measured. However, this distinction could be difficult to make, because many specimens included only the upper two feet of a plant at most. Also, at times it was difficult to distinguish between an axillary branch of a large inflorescence and the terminal portion of a stem (Figure 10). When available, a second leaf located three to five nodes below the inflorescence was measured in addition to the leaf directly below the inflorescence. 18

26 Georeferencing Geographic distribution is one characteristic used to define species or infraspecific taxa (du Rietz and Einar, 1930; Stuessy, 1990). Georeferencing and distribution mapping provide a way to visualize whether populations are sympatric, occupy different areas of a continuous range, or are geographically separated. Georeferencing also reveals whether differences in characters correspond to differences in geographic distribution. Only 14 of the 224 measured specimens listed geographic coordinates; 12 of those were specimens I collected in August Coordinates for these specimens were taken with a handheld geographic positioning system (GPS) unit. Coordinates for the collection localities of other specimens were found using the U.S. Geological Survey U.S. Board on Geographic Names (2010). Geographic coordinates could be found or estimated for 217 of the 224 specimens. Given that the range of data covered the entire eastern and Gulf coasts of the United States, localities were generalized; e.g., for example, coordinates for Ocean Springs were used for a location given as one mile east of Ocean Springs. For locations that were distant from a named location for instance, 35 airline miles NW of Miami or that involved the intersection of two roads, Google Earth was used (Google, 2009). Google Earth was also used to choose a roughly central point for very general locations, such as Maryland. Calvert County. Chesapeake Bay, or when only a county was 19

27 provided. The two collections made inland in Texas were collected at San Marcos Lake in Hays County, and at Grand Lake. Because there are two Grand Lakes in Texas, one in Stephens Co. and one in Montgomery Co., the latter was chosen as the probable location because it is closer to the coast. ArcMap software (ESRI Inc., ) was used to generate a map of geographic distribution. Geographic localities were categorized into four regions: the Gulf Coast (84 longitude and further west, including the Florida panhandle, = 1); the Florida peninsula (Florida, east of 84 longitude, = 2); the south Atlantic coast (North Carolina, South Carolina, and Georgia, = 3); and the mid-atlantic coast (Virginia and north, = 4). These categories provide shorthand to summarize the range of K. pentacarpos and to analyze patterns in character states across that range. Statistical Analyses A brief survey of revisions of single species shows that a variety of statistical methods can be used to analyze morphometric data. Statistical analysis methods that have been used for infraspecific revisions include principal components analysis, discriminant factor analysis, cluster analysis, analysis of variance (ANOVA), and Pearson s correlation. Principal components analysis is a method of statistical analysis intended to reduce the number of variables in a study to only those that are most important ( separating the signal from the noise ), and to recognize patterns in the relationships 20

28 between those variables. Its purpose is to reveal the most significant underlying trends in the data taken as a whole as a step in understanding the variance in the data. A principal components analysis of five recognized subspecies of Eriastrum densifolium (Polemoniaceae) (Brunell and Whitkus, 1998) found that character variation was too continuous to merit taxonomic distinction and recognition, with the exception of one subspecies, subsp. sanctorum, which is distinguished by corolla length. Their study utilized 196 herbarium specimens, measured for 19 quantitative characters and two qualitative characters (translated into numeric categories). All measurements were standardized using the STAND program, taxonomic differences between characters were calculated using the SIMINT program, character correlations were calculated using Statmost 3.0, and multivariate statistical analysis was performed using NTSYS-pc software. Cluster analysis was performed using UPGMA (Unweighted Pair Group Method with Arithmetic mean) to construct a dendrogram. Principal components analysis was performed using Eigenvalues and eigenvectors calculated using the EIGEN program. A morphometric study of Dioscorea quartiniana (Dioscoreaceae) (Wilkin, 1999) utilized principal components analysis to conclude that D. quartiniana is continuous in all characters and impossible to divide into subspecific taxa, thus eliminating dubious systems of varietal classification that had been used since To quantify the characteristics of 120 specimens of D. quartiniana, Wilkin took between one and five measurements for leaves and between four and 14 measurements for male flowers, female flowers, fruit, and seeds. The dioecious nature of the species required different measurements for male and female specimens; a maximum of 19 measurements were 21

29 taken for any one specimen. Wilken then calculated means and used logarithmic transformation to enhance representation of the magnitude of differences in the data. Principal components analysis showed that leaf shape differences are continuous rather than discrete. Male and female floral characteristics also formed a single cluster in multivariate space, indicating that there are no discrete character differences that would merit subspecific taxonomic recognition. Discriminant factor analysis, also known as multiple discriminant analysis and as canonical discriminant analysis, seeks to develop a rule that will accurately classify objects into different groups that are assumed to exist naturally. Canonical discriminant analysis is similar to analysis of variance, and was used in a revision of the North American Antennaria media complex (Asteraceae: Inuleae) (Chmielewski, 1997) in an attempt to produce a classification for this polyploid species complex that would be based on morphological characters and would not require knowledge of either chromosome count or reproductive history. Data were collected for 17 quantitative characters for more than 500 specimens. Length measurements were transformed to logarithms and count data were transformed to square roots, in order to give the data multivariate normality for the analysis. SAS DISCRIM and SAS UNIVARIATE software were used. Three data matrices were constructed to determine whether the specimens under consideration did in fact belong to the A. media complex, whether other putative taxa belonged in the complex, and whether the classification system held for both diploid and polyploid representatives. Discriminant function analysis was able to classify specimens correctly 22

30 more than 95% of the time for both diploid and polyploid specimens, upholding the existing subspecific classification system. A revision of Silene douglasii (Caryophyllaceae) (Kephart et al., 1999) used both principal components analysis and discriminant factor analysis to test several recognized varieties. Principal component analysis for 22 characters measured in 354 plants found identifiable clusters corresponding to recognized varieties douglasii, oraria, and rupinae. Principal component analysis was used to evaluate variance according to variety versus variance according to geography, and cohesive patterns were found for both. MINITAB software was used to conduct one-way analysis of variance. Discriminant factor analysis was used to test whether specimens could reliably be placed in one of the three putative varieties by morphological characteristics, to identify the most useful characters for identification, and to test whether altitudinal or habitat-related gradients were detectable within single populations. Morphological characters were sufficient to reliably distinguish the varieties. Most of the variation between specimens was shown to exist between varieties rather than within varieties. In the analysis of population differences, most variation was shown to exist between populations rather than within populations. Although discriminant factor analysis sometimes classified a specimen as belonging to the wrong population, it reliably classified the specimen as the correct variety. An infraspecific revision of the genus Margaretta (Asclepiadaceae) (Mwanyambo, 1996) used cluster analysis, analysis of variance, and canonical discriminant analysis. Morphometric data for 18 characters were collected from 100 herbarium specimens. Mwanyambo concluded that the genus can be defined into three 23

31 distinct groups, although not all characters partition distinctly into the three categories, possibly indicating infraspecific variation. Pearson s correlation is a measure of the linear dependence between two variables. Pearson s correlation was used to analyze clinal variation across the range of the Halesia carolina complex (Styracaceae) in the southeastern United States (Fritsch and Lucas, 2000). Measurements of 17 characters were taken from 286 herbarium specimens. The program SYSTAT version 5.2 was used to find the Pearson s correlation, to measure correlations between measured characters, and to measure variation across geographic clines. In addition, principal components analysis was used. The Pearson s correlations revealed significant associations between most (22 of 28, and 26 of 36) of the character pairs, indicating that variation in this species occurs gradually along a cline rather than discretely. It also found that many characters vary significantly according to latitude. The authors concluded that the H. carolina complex should be recognized as a single taxon without subspecific divisions. For the present infraspecific revision of Kosteletzkya pentacarpos, principal components analysis (PCA) and discriminant factor analysis (DA) were chosen to look for patterns in infraspecific variation and to test the reliability of recognition of the four varieties as delimited by published descriptions and floras. Pearson s correlation was used to aid in selecting characters for analysis in order to minimize correlation between characters. One-way analysis of variance (ANOVA) was run for each character measured against four designated geographic regions to test for correlation of character variation with geography. All analyses were run using SYSTAT Version 13 (SYSTAT, 2009). 24

32 For principal components analysis, the correlation matrix was used rather than raw covariance, because the ranges of the values were disparate, and because categorical and binary variables (e.g., pubescence) were included as well as continuous variables. The data did not have to be standardized prior to PCA because when the correlation matrix is used in SYSTAT data are standardized automatically using standard deviation. Components with eigenvalues of greater than one were retained, according to the latent root criterion (McGarigal et al., 2000). These components were saved, and the factor scores of the first and second component were graphed as a scatter plot. Prior to the discriminant analysis, the data were standardized to have a mean of zero and a standard deviation of one. The option compute from group sizes was chosen as opposed to all groups even, because the number of specimens of variety smilacifolia was much smaller than the number of specimens of each other variety. The phenology of the herbarium specimens varied: some were fruiting, others were flowering, some had both flowers and fruits, and some were sterile. Because of missing data, the data were analyzed in two subsets in order to maximize the number of specimens that could be included in each analysis. One subset excluded fruit characters to analyze specimens in flower, the other excluded flower characters to analyze specimens in fruit. Specimens having both flowers and fruit were included in both subsets. 25

33 Results Character Selection In total, 223 specimens had enough characters of sufficient quality for analysis, and were measured: 7 specimens of variety smilacifolia, 62 of variety aquilonia, 64 of variety althaeifolia, and 90 of variety virginica (Table 4). The preliminary list of morphological characters to measure was drawn from published descriptions of the varieties and from diagnostic characters used in floras, as well as a few characters that seemed intuitively useful. Early in the process of measurement, some characters that were initially recorded proved to be impractical and were excluded. The height of the plant was stated on the label on 157 sheets, but it was often rounded to the nearest foot, half-meter, or even meter, and so was inaccurate and unsuitable for statistical analysis. For 36 specimens, the whole plant was included and so the height could be accurately measured, but this number was too few to utilize in analysis. A character called petiole leaflet referred to the presence of a bud or cluster of small leaves on a petiole at a point that was not the node and was chosen based on initial examination of a few specimens. This character was excluded from the final analysis because it was not consistent for all petioles on a single plant. In addition to the leaf immediately below the inflorescence, the dimensions of a 26

34 second leaf, three to five nodes below the first, was also measured. However, a lower leaf was only available for measurement in 54 of the specimens measured, and so the lower leaf measurements were excluded. One flora, A Flora of Tropical Florida (Long and Lakela, 1971), described variety althaeifolia as having stinging hairs on the leaves and stems. However, only one specimen label mentioned stinging hairs: Crosby 1395 (DUKE), and no apparently urticacious hairs were seen under the microscope. A Flora of the Southeastern U.S. (Small and Kunkel, 1903) described the seeds of varieties virginica and smilacifolia as glabrous or obscurely wooly and seeds of variety althaeifolia as glabrous, striate-lineolate. However, it became evident that all mature seeds from all varieties were striate-lineolate, and no wooly seeds were seen in any specimen. Fourteen characters were highly correlated with other characters (Pearson s correlation r > 0.7), and so they were averaged or excluded in order to reduce statistical noise. Initially, both the longest petiole and the petiole with the most mature fruit were measured, but these were often one and the same petiole, and as a result they were highly correlated (r = for the petioles and also for the leaflets). The two pairs of measurements were averaged to produce a single value for pedicel length (PEDICEL) and one for the length of the subtending leaflet (SUBTLF). Longitudinal seed diameter (from the scar) and latitudinal seed diameter were both measured, but the seeds are generally round and these two measurements were highly correlated (r = 0.741). They were averaged to produce one value to represent the size of the seed (SEED). Hair length proved to be consistent on the abaxial and the adaxial side of a leaf (r = 0.921), and also on the high and medium portions of a stem (r = 0.889). Density was also highly 27

35 correlated between the abaxial and adaxial sides of a leaf (r = 0.802) but not between the high and middle portions of a stem (r = 0.631). Hair length on both sides of the leaf was averaged (LPUBLENG) as was density on both sides of the leaf (LPUBDENS). Hair length on the stem was also averaged (STMPUBLEN), but hair density on the stem was kept as two separate values. Petal length was highly correlated to both petal width (r = 0.771) and column length (r = 0.797). Petal length was retained to represent overall flower size, and petal width was retained so as not to lose information about the shape of the petals, but column length was excluded. The number of flowers and the number of lateral branches in the inflorescence, both representing inflorescence size, were highly correlated (r = 0.808). Number of lateral branches (BRANCH) was retained because it had more explanatory power than number of flowers in principal components analysis. Leaf width was highly correlated to several other leaf measurements: petiole length (r = 0.778), central vein length (r = 0.741), the distance from the petiole to the sinus (r = 0.783), and the width of the leaf between the sinuses (r = 0.807); all of these measurements were in part dependent on the overall size of the leaf. Because these other measurements were not as highly correlated to each other, leaf width was excluded from analysis. Ultimately, 33 characters were used in analyses (Table 2). Box-and-whisker plots of the ranges of quantitative characters provide an overview of the variation within each variety (Figures 11 18). Most characters show considerable overlap in ranges between the varieties, but a few show clear differences in means. Variety smilacifolia has shorter inflorescences (INFLOR), more slender stems (STEMDIAM), shorter leaf petioles 28

36 (LEAFPET), and smaller, narrower leaves (PETSIN, SINUSWID) than any other variety, as well as uniquely extremely reflexed lobes (ANGLE). Variety aquilonia has more constricted capsules than other varieties: while CAPDIAMLONG (representing the overall size of the capsule) is relatively constant across all varieties, CAPDIAMSHOR (width of capsule at constrictions) is clearly smaller. Seed size is generally largest in var. aquilonia, although its range overlaps with that for var. virginica. Principal Components Analysis Owing to phenology (in flower vs. in fruit) and missing data (Table 4), the data had to be divided into subsets and analyzed in two batches: one including flower characters (131 entities, 28 attributes), and one including fruit characters (107 entities, 31 attributes). These batches were run to maximize the number of specimens that could be included in each analysis; 75 specimens had both flower and fruit data and so were included in both batches. Principal components analysis was first run on a data set including flower characters but excluding fruit characters (131 entities, 28 attributes). The first component explains % of the total variance and the second %. Highly significant component loadings (> 0.50 ) (McGarigal et al., 2000) for the first component include INFLOR (0.643), BRANCH (0.639), STEMDIAM (0.794), LEAFPET (0.728), CENTVEIN (0.797), LOBE (0.520), PETSIN (0.877), SINUSWID (0.823), and TEETH (0.526). This suggests that the size of the inflorescence and the size of the leaves are of 29

37 primary importance in explaining the variation in the data set. Significant loadings for the second component include EPICLEN (0.551), SEPALLEN (0.528), SEPALWID (0.519), LPUBLENG (0.836), LPUBDENS (0.823), and STMPUBLENG (0.572). Pubescence, one of the most important characters in identifying the varieties in the field, is significant in explaining the variation in the data. So is flower size; calyx and epicalyx size are very important, and petal size is also important (PETALLEN = 0.495, PETALWID = 0.473). Overall, the points representing the specimens do not form discrete groups in multivariate space, although points representing var. aquilonia and points representing var. althaeifolia tend toward opposite sides of the graph (Figure 18). PCA run on the data set including fruit and excluding flowers (107 entities, 31 attributes) returned similar results. The first component explains % of the variation, and the second %. Highly significant loadings on the first component again mostly represent characters describing the size of the plant and leaves: INFLOR (0.582), BRANCH (0.579), STEMDIAM (0.705), LEAFPET (0.731), CENTVEIN (0.788), LOBE (0.613), PETSIN (0.875), SINUSWID (0.842), and TEETH (0.523). Significant loadings on the second component again include pubescence and calyx size EPICLEN (0.639), SEPALLEN (0.516), LPUBLENG (0.802), LPUBDENS (0.817), and STMPUBLENG (0.632). Pubescence on the fruit is also highly significant: CAPHAIRLEN (0.695), CAPHAIRDEN (0.626). Again, the scatterplot of factors 1 and 2 does not show discrete groups in multivariate space, although var. althaeifolia and var. aquilonia show tendency toward opposite extremes (Figure 19). 30

38 Discriminant Analysis A discriminant analysis of the flower data set classifies variety althaeifolia correctly at 85%, variety aquilonia at 90%, variety smilacifolia at 100%, and variety virginica at 73%. A jackknifed classification correctly classifies althaeifolia at 79%, aquilonia at 66%, smilacifolia at 75%, and virginica at 58%. Wilks lambda is significant (p < 0.001). The most important characters in discriminating between the groups are TEETH (f-to-remove = 7.887), LPUBDENS (7.154), EPICLEN (3.416), STMHIDENS (3.223), ANGLE (2.596), and SEPALWID (2.265). For the data set using fruit characters, discriminant analysis correctly classifies variety althaeifolia at 90%, aquilonia at 95%, smilacifolia at 100%, and virginica at 78%. The jackknifed classification correctly classifies althaeifolia at 61%, aquilonia at 62%, smilacifolia at 80%, and virginica at 58%. Wilks lambda is significant (p < 0.001). The most important characters in discriminating among groups are CAPHAIRLENG (10.722), TEETH (6.402), SEED (6.037), SEPALTIP (3.606), ANGLE (2.590), LEAFPET (2.327), STMPUBLENG (2.319), and STMMEDDENS (2.207). While variety smilacifolia forms a clearly delimited and discrete group based on the first two factors, the other three varieties do not show clear separation in either the discriminant analysis including flower characters or the analysis including fruit characters (Figures 20, 21). Instead, they appear to form a gradation based on the first and second factors; while varieties aquilonia and althaeifolia show more or less separation, variety virginica falls in the middle and overlaps them both considerably. This pattern echoes the 31

39 scatter plot of the first two factors from PCA. As in PCA, the first factor accounts for the most variation in the data, with the second and third and subsequent factors accounting for progressively less of the variation. Analysis of Variance by Geography Generally, specimens identified as variety althaeifolia are distributed in Virginia and to the south, specimens of var. aquilonia are distributed in North Carolina and to the north, and var. smilacifolia is restricted to the Florida peninsula. The typical variety is distributed throughout the range of the species (Figure 22). The range of K. pentacarpos was categorized into four regions: the Gulf Coast (84º longitude and further west, including the Florida panhandle, = 1); the Florida peninsula (Florida, east of 84º longitude, = 2); the south Atlantic (North Carolina, South Carolina, and Georgia, = 3); and the mid-atlantic (Virginia and north, = 4). One-way analysis of variance (ANOVA) was run for each morphological character against each region. Five characters were very significantly correlated (p < 0.001) with geographic region: length of capsule hairs (CAPHAIRLENG; f = ), density of capsule hairs (CAPHAIRDEN; f = ), density of leaf pubescence (LPUBDENS; f = 8.144), length of leaf pubescence (LPUBLENG; f = 7.100), and whether the leaf margin was serrate, biserrate, or entire (MARGIN; f = ). Two additional characters were also 32

40 significantly correlated (p < 0.01) with geographic region: seed size (SEED; p = 0.003, f = 2.240) and capsule constriction (CAPDIAMSHOR; p = 0.006, f = 3.862). Treatments in floras have used pubescence on leaves and capsules to differentiate varieties: the northern variety aquilonia has short, sparse pubescence on leaves and capsules, while the southern var. althaeifolia has long, dense pubescence. The ANOVA test and the distribution map shows that these are not distinct character states that define discrete populations, but instead are extremes of variation on a gradient across the species range. 33

41 Discussion Classification of Infraspecific Variation The International Code of Botanical Nomenclature (McNeill et al., 2006) formally recognizes five ranks underneath species: subspecies, variety, subvariety, form, and subform (Chapter I, article 4). Of these, three ranks are commonly used: subspecies, variety, and form (Stuessy, 1990). Use of the term variety for subspecific divisions and the practice of using categories below the species level for plants dates to Linnaeus Philosophica Botanica of Linnaeus used the term to denote accidental changes caused by environmental conditions or stresses. Today, we would say that his varieties do not hold true genetically; his use of variety is similar to modern use of the rank form (Stuessy, 1990). The term subspecies came into use in botany in the late 1700s, possibly from prior use in zoology. The term variety was used over subspecies by Asa Gray and A.P. de Candolle in the 1800s, and referred to morphologically or geographically distinct groups within a species (Stuessy, 1990). In this century, subspecies and variety have been used to refer to respectively broader and narrower morphological and geographic variation within a species. According to the widely followed definitions of du Rietz and Einar (1930), a subspecies 34

42 is a population of several biotypes forming a more or less distinct regional facies of a species, while a variety is a population of one or several biotypes, forming a more or less distinct local facies of a species [emphasis added]. Biotypes, by their definition, are groups of individuals with identical genotypes. In contrast to species, which he defines as the smallest natural populations permanently separated from each other by a distinct discontinuity, subspecies and varieties can cross more or less freely at areas of contact. During the early 1900s two schools of thought emerged in the United States: the Californian school considered the term subspecies to be correct, while the Eastern School considered variety to be correct. As a result, the two terms were often used interchangeably. At the 1975 International Botanical Congress in Leningrad there was a proposal to make the two terms equivalent, but it was not adopted. The use of both subspecies and variety to denote any grouping above form and below species was noted and lamented by de Rietz and Einar (1930), who complained that botanists were using the term [subspecies] in a rather chaotic way, and proposed that taxonomists either use the term variety to mean strictly local groupings, or abandon the term completely. Currently, some taxonomists still choose to use subspecies, some use variety, and some use both terms in hierarchy (Hamilton and Reichard, 1992). When they are used in hierarchy, subspecies are considered to be groups of similar varieties, or alternatively, varieties are considered to be distinct groups within a subspecies (Stuessy, 1990). Whether the infraspecific rank is called subspecies or variety, in current use it refers to morphologically and/or geographically distinct groups of populations within a 35

43 single species. In addition to external morphological attributes, patterns of variation in chemicals including secondary compounds and proteins can be used in classification (Smith, 1986) (Mwanyambo, 1996), as can ploidy levels, and even cultural uses in the case of domesticated species (Huaman and Spooner, 2002). While genetic divergence, the ability to produce hybrids, and the likelihood of hybrids occurring spontaneously in nature are also useful criteria in defining infraspecific taxa, these characteristics cannot be learned from herbarium specimens. Geographical distribution is an important part of determining if infraspecific taxa exist, because populations that are sympatric but always morphologically distinct could be assumed to be unable to hybridize readily, and so might be better defined as species (Stuessy, 1990). The infraspecific category form refers to distinctive phenotypes that occur sporadically in populations, and are not correlated with geography. A form may occur in different subspecies of a species (du Rietz and Einar, 1930). Plant taxonomists do not commonly formally recognize forms because these differences are minor, may be based solely on environmental conditions, and may co-exist with other forms in a single population, as opposed to the consistent and geographically linked patterns of morphological variation that distinguish species or varieties. In Kosteletzkya pentacarpos, one form has been published: K. virginica var. aquilonia forma alba, distinguished from variety aquilonia only in having a white flower (Reed, 1951). While Reed s publication of the form states that it was collected in Dorchester and Anne Arundel counties in Maryland, the loans I received contained seven whiteflowered specimens collected outside of Maryland. C.L. Pollard 1164 (duplicate from 36

44 NY) was collected in 1896 in Jackson Co., Mississippi; R.R. Smith et al. 654 (NCU) was collected in Dade Co., Florida in 1961; S. Vanderplank and M. Fishbein A (BRIT) was collected in Broward Co., Florida in 2002; S. Vanderplank 17B (BH) was collected in Broward Co., Florida in 2002; W.G. D Arcy 3050 (FLAS) was collected in Palm Beach Co., Florida in 1968; J.H. Willson s.n. (20 July 1967) (FLAS) was collected in Polk Co., Florida; and J.R. Orsegnio s.n. (25 July 1985) (FLAS) was collected in Palm Beach Co., Florida. I only considered a flower to be white if its label explicitly said that it was white, assuming that flowers that seemed white on the herbarium sheet may have faded through drying or storage. On my 2009 collecting trip through Maryland and Delaware I searched for but did not find any white flowers in Dorchester and Anne Arundel counties. The location at which Reed collected his type for forma alba has since been developed into residential housing. Recognizing Infraspecific Variation Formally naming infraspecific groups has been contentious, and the question of when to recognize subspecific groups is also problematic. Taxonomic distinctions made at one time based on a subset of a species may not hold true after a thorough, specimenbased review of the entire species. In their revision of Eriastrum densifolium (Polemoniaceae), Brunell and Whitkus (1998) found that many characters that had previously been assigned to one subspecific taxon were actually inconsistently present in many taxa. Furthermore, they found no discontinuities for any character between putative 37

45 infraspecific taxa, with the exception of one variety that could be consistently distinguished because of its distinctly longer corollas. Finally, it appeared that much of the variation was local, and could be attributed to ecological conditions in microgeographic areas (i.e., a single hill or valley). Because varieties and subspecies are not reproductively isolated as species are, theoretically, they may interbreed freely at transition zones, resulting in clinal patterns of variation over a geographic area rather than easily distinguished groups. Ideally, morphological differences between putative infraspecific groups should not form a continuous gradation, or even simply encompass different ranges, but should show disjunctions. As Fritsch and Lucas (2000) stated in their revision of the Halesia caroliniana complex (Styracaceae), just because means differ does not imply that there exists a distinct gap between the presumptive groups; all intermediates might well exist. They declined to recognize any infraspecific taxa within H. caroliniana because principal components analysis showed patterns of gradual clinal variation rather than distinct groupings; these clines largely corresponded to latitude. Even when discontinuities in morphology are apparent, they may not vary in consistent ways. An analysis of Eurasian Silene latifolia Poir. (Prentice, 1986) found that clustering and ordination analysis supported division into two groups, but that the zones of transition for individual characters did not coincide; that is, pollen morphology divided the populations into western and eastern groups, while the presence of genes coding for certain flavonoids divided the populations into north-western and south-eastern groups. 38

46 This incongruence also shows that visible, easily measured morphological characters may not vary in the same ways as chemical or cytological characters. The Chinese Paeonia delavayi complex had been variously interpreted as consisting of two species, one species with three infraspecific taxa, or three species with two infraspecific taxa. The different interpretations had been made based on various characters, including flower morphology and color and leaf morphology. The Hong et al. (1998) revision of the complex found a high level of variation in flower and leaf morphology within populations as well as between populations, an absence of correlation between the characters, and an absence of correlation with geography. Citing tremendous and continuous variation, they concluded that only one species without infraspecific taxa should be recognized. Even if subspecific classification is problematic, it is still useful. Stace (1986) provides a synopsis: Formal names provide convenient shorthand to refer to groups of populations, rather than having to resort to lengthy descriptions of characteristics and geographical areas. Naming a taxon draws attention to it. This can be especially important for directing research or conservation efforts. There may be groups under the level of species that are especially important for ecological, economic, research, or cultural reasons. If infraspecific ranks are not recognized, there is a danger of either recognizing too many minor variants as species and thereby making floras too long and 39

47 complicated, or neglecting to recognize them and thereby overlooking important information. Regarding the second item, recognizing subspecific variation can either help or hinder conservation efforts. Formally recognizing genetic variation by classifying that variation into formally named groups provides direction for preserving that variation. In addition, if an infraspecific taxon is rare or occupies a restricted area, it may merit conservation efforts in a way that the full species or a species with a larger circumscription does not. In 2003 Pino and de Roa stated, [K. pentacarpos] can be considered globally threatened because of its low number of localities. Indeed, in the European Union, K. pentacarpos appears only in a few scattered locations, and so is designated as deserving special areas of conservation (an Annex II species) (Pino and de Roa, 2003). However, because the common New World species K. virginica and the Old World K. pentacarpos have been determined to be a single species (Blanchard, 2008), K. pentacarpos can no longer be considered globally rare. It remains to be seen what impact this determination may have on the species conservation status. On the other hand, concentrating attention on recognizing ever-smaller taxonomic divisions can distract attention from distinctions that would be truly helpful for conservation purposes, since subspecific taxa are not reproductively isolated, theoretically, and so do not necessarily represent significantly different evolutionary lineages as do species. This may be the case in the orchid family, where continuous taxonomic attention to European taxa may be causing the taxonomy of tropical and neotropical orchids to be neglected. Also, checklists that must include large numbers of 40

48 similar taxa or large numbers of names that are no longer accepted are more difficult to compile (Pillon and Chase, 2006). Checklists are an integral part of defining and quantifying diversity, which is often the first step in conservation efforts. The third item the special importance of subspecific groups is especially true for cultivated plants, when characters such as sugar content or seed yield are of utmost importance and narrow classification is needed for setting quality control guidelines and prices. It is also true for wild species that may be sources of genetic variation for cultivated plants. Delimiting a species full range of morphological variation, as well as the ways in which its genetic variation corresponds to regional and environmental factors, can be valuable in optimizing the crop s yield and quality. A detailed revision of a species morphological variation can enable the compilation of a complete germplasm record and aid in identifying wild strains of the species that might be used to introduce useful genes into domesticated populations (Okafor, 1981). The definition of species has long been in debate, and there are still diverse opinions on what constitutes a species. Varieties and subspecies, that represent groups of populations that are not as genetically differentiated as species or groups of populations that are in the process of diverging to form new species, are also difficult to define formally. Generally, formally recognized infraspecific taxa should be geographically distinct and should differ consistently, with disjunctions, in multiple characters. Their variation should not be completely dependent on environmental factors, but should be genetically based, and the variation should not be sporadic as with the infraspecific category forma. 41

49 Infraspecific variation in Kosteletzkya pentacarpos In his treatment of Kosteletzkya pentacarpos for the Flora of North America North of Mexico (Volume 6, in press), Orland Blanchard decided to not formally recognize any infraspecific categories. His decision was based on the lack of both disjunctions between character ranges for the putative taxa and covariance between characters. In his words, while there may be some clinal variation in some characters, they do not vary together, with the result that no useful boundaries can be discerned (personal communication, 25 February 2008). The present morphometric and statistical review of the species has largely upheld his conclusion, with an exception. While it is evident that the variation between varieties althaeifolia, aquilonia, and virginica is continuous and largely clinal, variety smilacifolia exhibits clear separation in multivariate space and deserves formal recognition. Some diagnostic characters used in floras proved not to be useful in delimiting the putative taxa in Kosteletzkya pentacarpos. Fernald (1941) used pedicel diameter as a diagnostic character in his publication of variety aquilonia, but the ranges of pedicel diameter overlap considerably between the taxa (Figure 12). Fernald also stated that the pedicels of variety althaeifolia are shorter than their subtending leaves, while the pedicels of varieties aquilonia and virginica are longer than their subtending leaves. In fact, in the majority of specimens from all three varieties, pedicels are shorter than their subtending leaves. Shorter and longer pedicels occur in all three varieties, and the ratio of shorter vs. 42

50 longer pedicels is close: of the specimens measured for this study, 86% of althaeifolia specimens (56 of 65) had shorter pedicels, compared with 81% of aquilonia (50 of 62) and 77% of virginica (69 of 89). Small and Kunkel (1903) described the seeds of var. althaeifolia as striatelineolate and the seeds of the typical variety and var. smilacifolia as glabrous or obscurely wooly. It became evident through observation that all mature seeds were striate-lineolate regardless of taxon; some seeds were simply glabrous, but these appeared to be immature as they became flattened during the drying process. While pubescence on the fruit and on the herbage has often been used as a diagnostic character, especially to define variety althaeifolia, ANOVA analysis indicates that these characters follow a latitudinal cline. Northern populations generally have short, sparse pubescence on both fruit and herbage, while long dense pubescence is characteristic of plants from the Carolinas, Georgia, Florida, and around the Gulf Coast. Press (1999) provided a list of possible adaptive benefits of leaf pubescence, including protection from radiation from excessive sunlight, reduced water loss through trapped moisture and increased gas diffusion, and defense against predators such as insects. Hairier plants would then have an adaptive advantage in the stronger sunlight and longer seasons of insect activity in the southern regions of the United States. Seed size, leaf margin, and capsule constriction also vary along a roughly latitudinal cline. Northern populations generally have larger seeds than southern populations, biserrate margins as opposed to serrate leaf margins, and less tightly constricted capsules. Possibly the larger seeds of the northern populations are an adaptation to longer winters, although it is not 43

51 known how the thickness of the seed coat and amount of endosperm in the larger seeds compares with that of the smaller seeds. Continuing ecological research on Kosteletzkya pentacarpos at the University of Delaware may discover the ecological significance of increased pubescence and seed size for the plant, but has not yet done so (Jack Gallagher, personal communication 27 April 2010). The distribution of Kosteletzkya pentacarpos does not show geographic disjunctions but instead is continuous along the east and Gulf coasts (Figure 22). Plants conforming to the description of the typical variety are interspersed with populations of other varieties almost throughout the range. Some plants have characters that suggest one variety but fall within the range of a different variety. Commons s.n. (10 Sep 1918; PH) resembles var. althaeifolia in characters but was collected in Delaware. Three specimens that have the short sparse pubescence of var. aquilonia Demaree (NY), McDaniel (BRIT), and Bell 4679 (NCU) were collected respectively in Louisiana, Mississippi, and South Carolina. In addition, there are specimens that have characteristics of more than one putative taxon. Of the specimens analyzed, two that had been determined as variety althaeifolia also had the small flowers typical of variety aquilonia: Biltmore 3270 (GH) and Lovett 217 (USCH). Three additional specimens that were not identified to variety had both the small flowers of var. aquilonia and the heavy pubescence of var. althaeifolia: Cassen 409 (FLAS), Lowell s.n. (GH), and Wilbur 9228 (DUKE). 44

52 Evidently, varieties althaeifolia and aquilonia would be more accurately described as extremes in variation along a gradient rather than distinct varieties. Their distribution in multivariate space (Figures 18, 19, 20, and 21) supports this conclusion. Variety smilacifolia Variety smilacifolia, however, shows distinct separation from the other taxa in discriminant analysis of morphological characters (Figures 20, 21). It is distinctly smaller and less robust than the other varieties with a few-flowered and generally unbranched inflorescence (see boxplots for INFLOR and BRANCH, Figure 11) and a thin stem (STEMDIAM, Figure 16). The taxon has narrow sepals, tightly constricted capsules, and small seeds (Figure 12, Figure 14, Figure 15). Its most distinctive feature is its leaves: they are extremely narrow, some even constricting to a width of a couple of millimeters at the base of the blade, and the similarly linear lobes are reflexed more than 90 (Figure 17). The leaf margins are mostly entire, having teeth only in the distal half or third. Even given the wide variation of leaf shape and size within the species, the leaves of var. smilacifolia are immediately recognizable (Figure 8; Figure 23). These findings conform to Chapman s original publication of the variety, in which he described the stem as slender and the inflorescence as few-flowered. His first mention of the variety (1860) described the leaves as all hastate, with lanceolate serrate lobes, while his formal publication of the variety (1883) amended the description to note that the lowest leaves can be ovate and the leaf blade is serrate (Appendix C). 45

53 Variety smilacifolia s distribution is restricted to Florida. All specimens determined as variety smilacifolia were collected in Florida, with the exception of Pollard 1164 (BH, GH, NY) that did not conform to published descriptions of variety smilacifolia and so was a mistaken determination. The original publication of the variety by Chapman (1860, 1883) described the variety s range as South Florida. Small and Kunkel s flora (1903) also gave the range as south Florida, while Steere (1968) stated the range was western peninsular Florida. However, the formal recognition of variety smilacifolia must be made with caveats. It is rarely collected: only 18 of 1,114 specimens received on loan were var. smilacifolia, and owing to duplication and the poor condition of some of these specimens, only nine could be used in analysis. Very few of them had been collected recently: E.G. Riley 90 was collected in 1990, O.E. Frye 41D was collected in 1946, A.S. Hitchcock 18 and 19 were collected in 1900, F. Rugel s.n. and 196 were collected in 1845, and Chapman s collection (s.n.) lacked a date but can be assumed to have been made in the 1800s. These specimens lacked precise locality data: Hitchcock s collections were made in Myers in Lee Co., Florida, Rugel s somewhere along the Manatee River of Florida, and Chapman s collection was made in south Florida. A survey of two additional herbaria in Florida from which loans were not originally requested did not yield many additional specimens of var. smilacifolia. Of 37 specimens in Fairchild Tropical Garden s Virtual Herbarium (2007, accessed 14 April 2010), only one appeared to be variety smilacifolia (R.O. Woodbury s.n., collected in Martin Co. in 1988). The password-secured section of the Virtual Herbarium maintained 46

54 by the National Park Service contained 24 additional specimens, only one of which appeared to be var. smilacifolia (R.G. Reimus 1192, collected in Everglades National Park in 1997). The University of South Florida s Herbarium Specimen Database (Wunderlin and Hansen, 2008) provides photos of 100 K. pentacarpos specimens and only one appears to be var. smilacifolia: E. Jensen and C. Olson OS0563, collected in Oscar Scherer State Park in A number of botanists knowledgeable about the flora of Florida were also contacted. While all were familiar with Kosteletzkya pentacarpos, had noted the variety of leaf forms in the species, and recalled seeing plants in the field that resembled var. smilacifolia, none recognized formal varieties in the species and none were able to recall whether the smilacifolia-like plants were found in any specific habitat or under any unusual conditions (personal correspondence, Keith Bradley, 12 April 2010; Jimi Sadle, 13 April 2010; John Hays, 13 April 2010; John Tobe, 19 April 2010). It is possible that the distinct phenotype of variety smilacifolia is a response or an adaptation to environmental conditions such as moisture and soil nutrient content. Leaf size tends to be positively correlated with soil fertility and moisture, including rainfall and humidity, and negatively correlated with irradiance (Givnish, 1987). Jack Gallagher (personal communication, 27 Apr 2010) states that in agricultural experiments at the University of Delaware, We definitely find a response in plant size with nutrient addition in the local populations. The question is whether the distinctive morphological characteristics of var. smilacifolia represent an environmentally induced expression of the same genotype as var. pentacarpos, or whether the differences are genetically based and 47

55 thus have evolutionary significance (Quinn, 1978). The degree of genetic variation could be tested by experimentally growing both varieties from seed under identical conditions. If the differences in gene expression persist, var. smilacifolia truly represents a distinct lineage. Despite these caveats, I propose that variety smilacifolia should be recognized formally. Because Kosteletzkya virginica is now a synonym of Kosteletzkya pentacarpos, a new nomenclatural combination must be made for the variety. The first valid publication of the epithet smilacifolia was made by Chapman (1860: 57). The basionym for the new combination is Kosteletzkya virginica (L.) C.Presl ex A.Gray var. smilacifolia Chapm. Thus, the new combination is Kosteletzkya pentacarpos (L.) Ledeb. var. smilacifolia (Chapm.) S.N. Alexander, comb. nov. The key to the varieties of North American Kosteletzkya pentacarpos is as follows: 1 Inflorescence unbranched. Leaves linear, <20mm wide, lobed; lobes reflexed, linear; margin without serrations for proximal half of length of blade; [peninsular Florida] var. smilacifolia 1 Inflorescence branched. Leaves hastate or lanceolate; >20mm wide; base cordate, truncate, or lobed, but lobes not reflexed; margin serrate or biserrate; [east and Gulf coasts including Florida] var. pentacarpos 48

56 Future research Because Kosteletzkya pentacarpos var. smilacifolia has been so infrequently collected, it should be intentionally collected to determine whether it is rare or whether it has simply been undercollected. Even though the species K. pentacarpos is common in the United States, if variety smilacifolia is indeed rare it may merit protected conservation status. As part of its range falls within Everglades National Park, it already has a measure of protection. Botanists collecting var. smilacifolia should take careful note of its geographic location and edaphic conditions to discover whether it is sympatric with the typical variety, whether it is adapted to a different niche, and whether it forms the dense populations that K. pentacarpos var. pentacarpos often forms. Ecological studies at the University of Delaware have found that K. pentacarpos plants grow larger in response to nutrient addition (J. Gallagher, personal correspondence April ), so it would be instructive to test the composition of the soil on which variety smilacifolia grows to test to what degree its small size might be a response or an adaptation to nutrient-poor soil. It would also be useful to collect fertile seeds, so garden plots could be grown to test the plant s response to various ecological conditions in a controlled setting, and to compare its responses to those of var. pentacarpos. Crossing studies could also be done to determine whether the varieties show any degree of reproductive isolation. Molecular genetic analysis could also determine whether var. smilacifolia is a single evolutionary 49

57 lineage and whether it has derived from within or has evolved as sister to var. pentacarpos. Clarifying the relationship between New World Kosteletzkya pentacarpos and Old World K. pentacarpos also remains to be determined. Orland Blanchard has speculated that K. pentacarpos may have been introduced into Europe in ship s ballast from the United States (personal communication, 24 September 2008) based on his observations that most early collections of European K. pentacarpos were made in or near major coastal port cities (including Valencia and Barcelona in Spain and Venice and Naples in Italy), and Old World populations are rare and scattered. A possible direction for future research would be a formal study of the relationship between New World and Old World K. pentacarpos using molecular genetic methods. If Blanchard s hypothesis is correct, then Old World K. pentacarpos may display a founder effect; a subset of the genetic diversity found in New World K. pentacarpos. Comparing the genetic diversity of New World and Old World K. pentacarpos could illuminate processes of speciation currently at work. 50

58 Appendix A: Tables Table 1. Sources of loans, total specimens received on loan, and total specimens measured. Acronym Institution State # Specimens Received AUA Auburn University Alabama 16 3 BH Cornell University New York BRIT Botanical Research Institute Texas of Texas DOV Delaware State University Delaware 4 0 DUKE Duke University North Carolina FLAS Florida Museum of Natural Florida History GA University of Georgia Georgia GH Gray Herbarium, Harvard University Massachusetts GMUF George Mason University Virginia 16* 14* NCU University of North Carolina North Carolina NY New York Botanical Garden New York ODU Old Dominion University Virginia 14 0 PH Academy of Natural Sciences Pennsylvania TEX University of Texas at Austin Texas USCH University of South Carolina South 21 8 Carolina TOTAL 1, *Includes 12 collections made by S. Alexander, A. Weeks, and K. Baird. # Specimens Measured 51

59 Table 2. Characters measured. Category Character Symbol (if used in analysis) Measurement Units or Qualities Vegetative Inflorescence length INFLOR Centimeters (rounded to halfcentimeter) Vegetative Number of lateral branches BRANCH Count in inflorescence Vegetative Number of flowers in Count inflorescence Vegetative Pedicel length (most mature Millimeters (whole) fruit) Vegetative Subtending leaflet length Millimeters (whole) (most mature fruit) Vegetative Pedicel length (longest) Millimeters (whole) Vegetative Subtending leaflet length Millimeters (whole) (longest) Vegetative Pedicel length (average of PEDICEL** Millimeters (whole) longest and most mature) Vegetative Subtending leaflet length SUBTLF** Millimeters (whole) (avg longest and most mature) Vegetative Petiole leaflet Presence/absence Flower Petal length PETALLEN Millimeters (whole) Flower Petal width PETALWID Millimeters (whole) Flower Column length Millimeters (whole) Calyx Pedicel diameter PEDDIAM Millimeters (fraction) Calyx Epicalyx length EPICLEN Millimeters (fraction) Calyx Epicalyx bract shape EPICSHAP Linear or flattened* Calyx Sepal length SEPALLEN Millimeters (fraction) Calyx Sepal width SEPALWID Millimeters (fraction) Calyx Sepal shape SEPALSHAP Lanceolate or ovate-lanceolate* Calyx Sepal tip shape SEPALTIP Acuminate or Fruit Long capsule diameter (ridge to ridge) CAPDIAMLONG acute* Millimeters (whole) 52

60 Category Character Symbol (if used in analysis) Measurement Units or Qualities Fruit Short capsule diameter CAPDIAMSHOR Millimeters (whole) (constriction to constriction) Fruit Capsule hair length CAPHAIRLEN <0.5mm; 0.5-1mm; >1mm* Fruit Capsule hair density CAPHAIRDEN >0.5mm apart; mm apart; <0.25mm apart* Fruit Seed diameter (longitudinal from scar) Millimeters (fraction) Fruit Seed diameter (latitudinal) Millimeters (fraction) Fruit Seed size (average of latitudinal and longitudinal SEED** Millimeters (fraction) diameters) Fruit Seed vestiture Glabrous, striate, wooly Leaf Stem diameter at upper leaf STEMDIAM Millimeter (rounded to halfmillimeter) Leaf Depth of base (if cordate or BASE Millimeters (whole) hastate) Leaf Petiole length LEAFPET Millimeters (whole) Leaf Length of blade (central CENTVEIN Millimeters (whole) vein) Leaf Length of petiole to lobe LOBE Millimeters (whole) Leaf Length of petiole to sinus PETSIN Millimeters (whole) Leaf Length of sinus to sinus SINUSWID Millimeters (whole) Leaf Angle of lobe from central ANGLE Degrees vein Leaf Total leaf width Millimeters (whole) Leaf Leaf margin MARGIN Entire; serrate; biserrate* Leaf Length that contains 5 teeth TEETH Millimeters (whole) Pubescence Abaxial leaf pubescence length Cinereous; hispid; hirsute* Pubescence Abaxial leaf pubescence Sparse; moderate; Pubescence density Adaxial leaf pubescence length dense* Cinereous; hispid; hirsute* 53

61 Category Character Symbol (if used in analysis) Measurement Units or Qualities Pubescence Adaxial leaf pubescence density Sparse; moderate; dense* Pubescence Leaf pubescence length (adaxial and abaxial LPUBLENG** Cinereous; hispid; hirsute* averaged) Pubescence Leaf pubescence density (adaxial and abaxial LPUBDENS** Sparse; moderate; dense* averaged) Pubescence High stem pubescence length Cinereous; hispid; hirsute* Pubescence High stem pubescence density STMHIDENS Sparse; moderate; dense* Pubescence Medium stem pubescence length Cinereous; hispid; hirsute* Pubescence Stem pubescence length (high and low averaged) STMPUBLEN** Cinereous; hispid; hirsute* Pubescence Medium stem pubescence density STMMEDLENG Sparse; moderate; dense* *See Table 3: Qualitative values key. ** Values for analysis calculated from measurements. 54

62 Table 3. Qualitative values key. Epicalyx bract shape Sepal shape Sepal tip shape Leaf margin Vegetative Pubescence Length Vegetative Pubescence Density Capsule Pubescence Length Capsule Pubescence Density Value Term 0 Linear 1 Flattened Definition as Used 0 Lanceolate Sides straight; shape more triangular 1 Ovatelanceolate Sides rounded 0 Acuminate 1 Acute 0 Entire 1 Serrate 2 Biserrate 1 cinereous Double arms <0.25mm and thin 2 hispid Single arm of hair <0.25mm, double arms >0.25mm 3 hirsute Single arm of hair >0.25mm 1 sparse Adjacent hairs do not overlap at all 2 moderate Hairs overlap at tips; individual stellate hairs can be distinguished 3 dense Hairs overlap heavily; individual stellate hairs difficult to distinguish 1 short Long hairs <0.5mm 2 medium Long hairs 0.5 1mm 3 long Long hairs >1mm 1 sparse Long hairs >0.5mm apart 2 moderate Long hairs mm apart 3 dense Long hairs <0.25mm apart 55

63 Table 4. Data coverage; characters present for specimens measured. smilacifolia aquilonia althaeifolia virginica TOTAL All characters present Flower, leaf, fruit, but no seed Flower and leaf Flower and full fruit Flower and fruit 1 1 without seed Leaf and full fruit Leaf and fruit without seed Only flower Only fruit (full) Only fruit 1 1 (without seed) Only leaf TOTAL

64 Appendix B: Figures Figure 1. Kosteletzkya pentacarpos (L.) Ledeb. flowers. Note color variation of petals and stigmatic surfaces. Bar = 1 cm. Photo by Andrea Weeks. 57

65 Figure 2. Kosteletzkya pentacarpos (L.) Ledeb. Typical habit and habitat. Bar = 50 cm. Photo by Andrea Weeks. 58

66 Figure 3. Range of Kosteletzkya pentacarpos in the United States. Image credit: U.S Department of Agriculture, PLANTS Profile ( Figure 4. Range of Kosteletzkya pentacarpos in Virginia. Image credit: Virginia Botanical Association Digital Atlas of the Virginia Flora ( 59

67 Figure 5. Hair length and density. A: short and sparse (Fernald (GH)); B: medium length and medium density (Demaree (GH)); C: long and dense (Wilbur (DUKE)). Millimeters shown on ruler. 60

68 Figure 6. Detail of leaf abaxial surface, showing 5-armed stellate hairs and large 3-armed stellate hairs (Ahles (BRIT)). Millimeters shown on ruler (left). Arrows indicate centers of 3-armed stellate hairs. 61

69 Figure 7. Fruit capsule constriction and pubescence in Kosteletzkya pentacarpos. A: a deeply constricted capsule with long hairs (Webster 4225 (GH)); B: a shallowly constricted capsule with short hairs (Alexander 10 (GMUF)). Millimeters shown on ruler (left). 62

70 Figure 8. Leaf shape variation in Kosteletzkya pentacarpos. A: hastate with reflexed lobes, margin entire (typical of variety smilacifolia) (Frye 31D (FLAS)); B: lanceolate with hastate lobes and serrate margins (Fernald (GH)); C: ovate and palmately lobed with biserrate margins (Curtiss 5626 (NY)). 63

71 Figure 9. Measurements taken to represent leaf dimensions for Kosteletzkya pentacarpos. 1: petiole length (LEAFPET), 2: depth of base (BASE), 3: length of blade (central vein) (CENTVEIN), 4: length base to lobe (LOBE), 5: length petiole to sinus (PETSIN), 6: length of sinus to sinus (SINUSWID), 7: angle of lobe from central vein (ANGLE), 8: total width (not used in analysis). 64

72 Figure 10. An axillary branch from A.H. Curtiss 5696 (GH) (to the left); note that it could be mistaken for the terminal portion of a small stem with a small inflorescence if cut differently. 65

73 Figure 11. Inflorescence character measurements for each putative taxon: length of inflorescence (INFLOR), number of lateral branches (BRANCH), average of length of longest flower/fruit pedicel and most mature pedicel (PEDICEL); average of length of subtending leaflet under longest and under most mature flower/fruit pedicel (SUBTLF). For all box-and-whisker plots, the whiskers indicate the extent of the inner fence, at the upper or lower quartile plus 1 ½ the inner quartile range. Asterisks indicate outliers within the outer fence, at the upper or lower quartile plus 3 the IQR; circles indicate outliers outside the outer fence. 66

74 Figure 12. Calyx character measurements for each putative taxon: pedicel (millimeters, PEDDIAM), length of epicalyx (mm, EPICLEN), length of sepal (mm, SEPALLEN), width of sepal (mm, SEPALWID). 67

75 Figure 13. Flower character measurements for each putative taxon: petal length (mm, PETALLEN), petal width (mm, PETALWID). 68

76 Figure 14. Fruit character measurements for each putative taxon: width of capsule at largest point of carpels (mm, CAPDIAMLONG), width of capsule at constrictions (mm, CAPDIAMSHOR). Figure 15. Seed sizes (average of length and width of round seed, mm, SEED) for each taxon. 69

77 Figure 16. Leaf character measurements for each putative taxon: diameter of stem at first leaf underneath inflorescence (millimeter, STEMDIAM), depth of base (if cordate or hastate) (mm, BASE), petiole length (mm, LEAFPET), length of blade (mm, CENTVEIN). 70

78 Figure 17. Leaf character measurements for each putative taxon: length of lobe (mm, LOBE), length of petiole to sinus (mm, PETSIN), width from sinus to sinus (mm, SINUSWID), angle between central vein of blade and lobe (degrees, ANGLE). 71

79 Figure 18. Principal components analysis flowering specimens; factor 1 against factor 2. A = variety althaeifolia, Q = variety aquilonia, S = variety smilacifolia, V = variety virginica. 72

80 Figure 19. Principal components analysis of fruiting specimens; factor 1 against factor 2. A = variety althaeifolia, Q = variety aquilonia, S = variety smilacifolia, V = variety virginica. 73

81 Figure 20. Discriminant analysis of flowering specimens. A = variety althaeifolia, Q = variety aquilonia, S = variety smilacifolia, V = variety virginica. 74

82 Figure 21. Discriminant analysis of fruiting specimens. A = variety althaeifolia, Q = variety aquilonia, S = variety smilacifolia, V = variety virginica. 75

83 Figure 22. Distribution of examined Kosteletzkya pentacarpos specimens on the east coast of the United States. = variety althaeifolia, = variety aquilonia, = variety smilacifolia, = variety virginica. 76

84 Figure 23. Representative specimen of Kosteletzkya pentacarpos var. smilacifolia. Frye 41D (FLAS). 77